CN112836274B - Data fusion method for tracing and auditing hidden engineering - Google Patents
Data fusion method for tracing and auditing hidden engineering Download PDFInfo
- Publication number
- CN112836274B CN112836274B CN202110141287.XA CN202110141287A CN112836274B CN 112836274 B CN112836274 B CN 112836274B CN 202110141287 A CN202110141287 A CN 202110141287A CN 112836274 B CN112836274 B CN 112836274B
- Authority
- CN
- China
- Prior art keywords
- defect
- detection signal
- data
- fusion
- composite detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Civil Engineering (AREA)
- Architecture (AREA)
- Probability & Statistics with Applications (AREA)
- Structural Engineering (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides a data fusion method for blind engineering tracking audit, belonging to the technical field of engineering audit. The method comprises the steps of detecting internal data of the hidden project by using various nondestructive testing sensors, establishing a composite testing signal defect characteristic database, then utilizing classification probability to keep a discriminant analysis algorithm to extract composite testing signal defect characteristics, carrying out self-adaptive real-time flow data fusion processing, and finally obtaining the internal damage type of the hidden project through a judgment result, so that constructors can conveniently know the internal conditions of the hidden project, and the hidden project can be better maintained. The data fusion method adopted by the invention gets rid of the limitation that the traditional data fusion is difficult to adapt to environmental changes; the invention also provides a parallel data fusion method, which avoids the performance delay of multi-sensor signal defect characteristic centralized task scheduling and improves the timeliness of system processing.
Description
Technical Field
The invention belongs to the technical field of engineering audit, and particularly relates to a data fusion method for concealed engineering tracking audit.
Background
With the rapid development of Chinese economy, a large number of regional or national basic construction projects such as public buildings, bridges, high-speed rails, water conservancy, expressways and the like are developed in succession, 2015 the investment of fixed assets in China reaches 2.5 ten thousand yuan, and the fixed assets in China are far ahead of other countries in the world. The hidden project is the foundation and important component of the whole public project, the importance of the public project audit is increasingly prominent, and the hidden project audit is the key point and the difficulty of the whole public project audit. The main difficulty of the concealed engineering tracking audit is to judge whether the materials, specifications, models, engineering quantities and the like used by the engineering are consistent with the construction design drawing or not, and whether the materials and the construction process of the concealed engineering can be continuously tracked and monitored or not.
Public works audit has gradually adopted detection means such as ultrasonic waves, electromagnetism, eddy currents, laser, geological radar, video monitoring, big dipper location to visualize the interior data of the hidden works, and a single sensor can detect partial interior state data of the hidden works, but the data are fragmented, so that the ability of revealing the whole interior data of the hidden works is lacked. When a plurality of sensors are adopted for detection, the traditional composite nondestructive detection data fusion method is roughly divided into two types: one is to directly perform pixel level fusion on the gray level of a source image according to a certain fusion rule; the second type is that the characteristic information of the image is extracted on the basis of the pixel level to carry out comprehensive analysis and fusion processing, the fusion of the characteristic level image can compress the information, the composite characteristic of the image can be kept, and the help can be directly provided for the fusion analysis of the decision level.
However, the existing data fusion method has the following defects:
(1) at present, aiming at the fusion processing aspect of specific signal defect characteristics, the research on the non-destructive detection data fusion algorithm of a double detector and a multi detector is more, but under the conditions of fixed detection objects and determined environmental conditions, the research on the self-adaptive fusion algorithm under the condition of dynamic detection target is less;
(2) at present, the research on the universal adaptive fusion algorithm is more, but the research on the adaptive fusion algorithm of a multi-detection sensor in the field of nondestructive detection needs to be further enhanced;
(3) from the perspective of high real-time fusion processing performance, the existing nondestructive testing method has no specific targeted research on the parallelization aspect of the fusion algorithm in the nondestructive testing field.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a data fusion method for the tracing and auditing of the hidden project, which establishes a typical signal defect characteristic database and further performs characteristic extraction and self-adaptive real-time flow data fusion processing, thereby effectively solving the limitation that the traditional data fusion method cannot adapt to environmental change, solving the problem of performance delay of centralized task scheduling of multi-sensor signal defect characteristics and improving the timeliness of system processing.
The present invention achieves the above-described object by the following technical means.
A data fusion method for tracing and auditing of covert engineering comprises the following steps:
step 1: detecting the internal state of the hidden project in real time by utilizing a plurality of nondestructive testing sensors, and transmitting the detected data to a central server;
step 2: dividing the detection task of the nondestructive detection sensor into local detection fusion and global detection fusion, and performing parallel fusion pretreatment;
and step 3: the central server receives the concealed project internal data detected by the nondestructive testing sensor, namely composite testing signal data, and carries out fusion processing on the composite testing signal data; the fusion processing process comprises the following steps: establishing a composite detection signal defect feature database, extracting composite detection signal defect features by using a classification probability maintenance discriminant analysis algorithm, and performing self-adaptive real-time flow data fusion processing on the extracted composite detection signal defect features;
and 4, step 4: the central server submits the fusion result generated in step 3 for the operator to view.
Further, in the step 1, the nondestructive testing sensor comprises an infrared thermal imaging device, an ultrasonic detector and an eddy current sensor, and is respectively used for detecting a near-surface image in the concealed project, a large-depth defect in the concealed project, a defect of a specific steel plate, and the diameter and the number of the steel bars.
Further, in the step 2, the specific process of the parallel fusion preprocessing is as follows: defining an initial RDD: and reading the composite detection signal data from the external data space into the system when the Spark program runs, and converting the composite detection signal data into Spark data blocks to form initial RDD.
Further, in the step 3, the specific process of establishing the composite detection signal defect feature database is as follows: the defect modeling and inputting module in the central server software is used for depicting and representing the defect characteristics of the composite detection signal, and a multi-dimensional classification model of the defect characteristics of the composite detection signal is established; constructing a composite detection signal defect feature vector matrix according to the multi-dimensional classification model, and simply classifying defect features; and simulating the defect characteristic data of the large-scale composite detection signal by using a simulator, and establishing a typical composite detection signal defect characteristic system and a typical composite detection signal defect characteristic database.
Further, in step 3, the specific process of extracting the defect feature of the composite detection signal is as follows: and taking the data in the composite detection signal defect feature database as original samples, calculating the classification probability of each sample to obtain a corresponding class center point and classification information, representing the physical distribution of the corresponding samples through the class center point and the classification information, obtaining more effective composite detection signal defect features representing the original samples, and extracting the composite detection signal defect features.
Further, in step 3, the specific process of performing adaptive real-time stream data fusion processing on the extracted defect features of the composite detection signal is as follows:
when m characteristics are fused, m groups of classifiers are needed for classificationThe probability output of the instrument is denoted as pij(class i | input), wherein i represents the i-th class decision tag, and j represents the j-th class of characteristics generated by the j-th group of classifiers as the decision result;
for the test sample, the probability output corresponding to the decision result of the jth group of classifiers is the maximum value of the probability output corresponding to all decision labels of the group of classifiers, and is recorded as pj：
Then merge weight wjComprises the following steps:
the extracted defect characteristics of the composite detection signal are used as test samples and input into a k-class PSVM classifier, k classes of defect characteristic signals to be trained and classified are set, and m types of defect characteristics of the extracted composite detection signal are set;
respectively inputting m different defect characteristics to be trained into m groups of classifiers, and training according to a tree structure, wherein each group of classifiers comprises 1 PSVM; when the input of the PSVM is more than two types, uniformly dividing the category labels into two categories for judgment until the input of the PSVM at the bottom layer is the two categories, and finishing the tree structure training;
respectively inputting m different defect characteristics to be tested into m groups of classifiers to obtain probability output p of all the classifiersij(class i | input), m groups of classifiers have m probability outputs, and then the fusion weight w corresponding to each group of classifiers is calculatedj；
Weighting the probability output corresponding to each decision label in each group of classifiers, wherein the weighting result is recorded as f (x), f (x) represents the class weight, and the calculation formula is as follows:
and outputting the judgment label corresponding to the maximum weighting result as a fusion result.
Further, the specific process of step 4 is as follows: and the central server utilizes a Spark application program to carry out conversion operation on the initial RDD to form a new RDD, triggers a Spark driver and submits a fusion result generated by the classifier.
The invention has the following beneficial effects:
the data fusion method provided by the invention characterizes the composite detection signal defect characteristics, establishes a multi-dimensional classification model of the composite defect characteristics, constructs a vector matrix, and can effectively reveal the modeling rule of the target defect characteristics, thereby establishing a typical signal defect characteristic database. The invention provides a classification probability retention discriminant analysis feature extraction method aiming at various defect features of high noise, has better feature extraction effect and provides convenience for the subsequent data fusion processing process.
The invention constructs and realizes the PSVM algorithm based on the signal defect characteristics, and gets rid of the limitation that the traditional data fusion is difficult to adapt to the environmental change; meanwhile, a parallel data fusion method is also provided, so that the performance delay of multi-sensor signal defect characteristic centralized task scheduling is avoided, and the timeliness of system processing is improved. The data fusion method for the blind engineering tracking audit greatly improves the precision and speed of nondestructive detection.
Drawings
FIG. 1 is a flow chart of a data fusion method according to the present invention.
Detailed Description
The invention will be further described with reference to the following figures and specific examples, but the scope of the invention is not limited thereto.
The data fusion method for the blind engineering tracking and auditing is shown in figure 1 and comprises the following steps:
step 1: detecting the internal state of the hidden project in real time by utilizing various nondestructive testing sensors, wherein the nondestructive testing sensors comprise infrared thermal imaging equipment, an ultrasonic detector, an eddy current sensor and the like and are respectively used for detecting data such as a near-surface image, an internal large-depth defect, a defect of a specific steel plate, the diameter and the number of reinforcing steel bars and the like in the hidden project; and data detected by the nondestructive testing sensor is transmitted to a data fusion module on the central server for data fusion processing so as to realize continuous tracking monitoring of the hidden project.
Step 2: in order to solve the problem of performance delay of task scheduling in a multi-detection sensor signal defect feature set and improve the timeliness of data processing, detection tasks of a nondestructive detection sensor are divided into local detection fusion and global detection fusion, and parallel fusion pretreatment is carried out: constructing a self-adaptive parallel computing cluster processing scheme based on memory optimization based on a cluster computing engine Spark; defining an initial elastic Distributed data set (RDD): and reading the composite detection signal data from the external data space into the system when the Spark program runs, and converting the composite detection signal data into Spark data blocks to form initial RDD.
And step 3: the central server receives various types of data, namely composite detection signal data, in the concealed engineering detected by the nondestructive detection sensor, and performs fusion processing on the composite detection signal data, wherein the specific processing process is as follows:
step 3.1: establishing a composite detection signal defect characteristic database;
the invention relates to a hidden project defect classification method, which comprises the steps of utilizing a defect modeling and inputting module in central server software to depict and characterize the defect characteristics of a composite detection signal and establishing a composite detection signal defect characteristic multi-dimensional classification model, wherein the defect types in the hidden project comprise hollowing, internal cracks, slag inclusion, diameter loss of reinforcing steel bars and the like; constructing a composite detection signal defect feature vector matrix according to the model, and simply classifying defect features; simulating the defect characteristic data of the large-scale composite detection signal by using a simulator, and establishing a typical composite detection signal defect characteristic system and a database;
step 3.2: extracting the defect characteristics of the composite detection signal;
because the number of original features in the composite detection signal defect feature database is large, a linear discriminant analysis method is needed to perform dimension reduction processing on the original features, and meanwhile, each category can be clearly reflected on low-dimensional data; in the linear discriminant analysis process, the distance between each sample vector and an intra-class mean vector is minimized, and the distance between each sample vector and an inter-class mean vector is maximized, so that dimension reduction processing is realized, wherein the sample vectors refer to sample data in a composite detection signal defect feature database, the intra-class mean vector refers to a random vector mathematical expectation of the same class of defects in the composite detection signal defect feature database, and the inter-class mean vector refers to a random vector mathematical expectation of different classes of defects in the composite detection signal defect feature database;
the linear discriminant analysis does not consider the distribution of the original samples, and is executed by a hard classification standard every time, so that the invention introduces classification probability to represent the distribution information of the samples in the linear discriminant analysis process, further provides a classification probability maintenance discriminant analysis (CPPDA) algorithm, and extracts the features in the composite detection signal defect feature database, and the specific process is as follows:
taking data in a composite detection signal defect feature database as original samples, calculating the classification probability of each sample to obtain a corresponding class center point and classification information, expressing the physical distribution of the corresponding samples through the class center point and the classification information, finally obtaining more effective composite detection signal defect features representing the original samples, and effectively extracting the composite detection signal defect features;
step 3.3: performing self-adaptive real-time flow data fusion processing on the composite detection signal defect characteristics extracted in the step 3.2;
according to the property of a traditional k classification Probability Support Vector Machine (PSVM), testing the probability that a sample belongs to a k label, wherein in the k classification problem, the higher the probability value corresponding to a judgment result is, the more accurate the judgment of a classifier is; when m characteristics are fused, m groups of classifiers are needed, and the probability output of the classifiers is recorded as
pij(class i | input) represents a composite weight matrix, wherei represents the ith class decision label, and j represents the j class characteristic generated by the j class classifier; for the test sample, the probability output corresponding to the decision result of the jth group of classifiers is the maximum value of the probability output corresponding to all decision labels of the group of classifiers, and is recorded as pjThe calculation formula is as follows:
pjthe larger the test sample is, the more accurate the classification judgment of the group of classifiers on the test sample is;
when multi-feature classifier fusion is considered, the significance of the weight is the importance of each set of classifier judgment results in the final fusion result; in the classification problem, the judgment accuracy of the classifier can be used as an index for describing the importance, and the fusion weight w provided by the inventionjThe calculation formula of (a) is as follows:
taking the various composite detection signal defect characteristics extracted in the step 3.2 as test samples, inputting the test samples into a classifier, setting k types of defect characteristic signals to be trained and classified, and m types of extracted composite detection signal defect characteristics;
respectively inputting m different defect characteristics to be trained into m groups of classifiers, and training according to a tree structure, wherein each group of classifiers comprises 1 PSVM; when the input of the PSVM is more than two types, uniformly dividing the category labels into two categories for judgment until the input of the PSVM at the bottom layer is the two categories, and finishing the tree structure training;
respectively inputting m different defect characteristics to be tested into m groups of classifiers to obtain probability outputs of all the classifiers, wherein the m groups of classifiers have m probability outputs; calculating the fusion weight w corresponding to each group of classifiers according to the formulas (1) and (2)j(ii) a The probability corresponding to each decision label in each group of classifiers is inputAll the outputs are weighted, the weighting result is marked as f (x), f (x) represents the category weight, and the calculation formula is as follows:
and finally, outputting the judgment label corresponding to the maximum weighting result as a fusion result, wherein the fusion result is the final judgment result of the classifier.
And 4, step 4: and (3) the central server performs conversion operation on the initial RDD according to the Spark application program defined in the step (2) to form a new RDD, triggers a Spark driver, and submits a judgment result generated by the classifier in the step (3) for a worker to view. According to the final judgment result of the classifier, the specific damage type in the hidden project can be obtained, so that the construction personnel can conveniently know the internal condition of the hidden project, and the hidden project can be better maintained. The PSVM algorithm based on the signal defect characteristics is a self-adaptive fusion algorithm of multiple detection sensors in the field of nondestructive detection, can adapt to detection under the condition of a dynamic detection target, and gets rid of the limitation that the traditional data fusion algorithm is difficult to adapt to environmental changes.
The present invention is not limited to the above-described embodiments, and any obvious improvements, substitutions or modifications can be made by those skilled in the art without departing from the spirit of the present invention.
Claims (3)
1. A data fusion method for tracing and auditing of covert engineering is characterized by comprising the following steps:
step 1: detecting the internal state of the hidden project in real time by utilizing a plurality of nondestructive testing sensors, and transmitting the detected data to a central server;
step 2: dividing the detection task of the nondestructive detection sensor into local detection fusion and global detection fusion, and performing parallel fusion pretreatment;
and step 3: the central server receives the concealed project internal data detected by the nondestructive testing sensor, namely composite testing signal data, and carries out fusion processing on the composite testing signal data; the fusion processing process comprises the following steps: establishing a composite detection signal defect feature database, extracting composite detection signal defect features by using a classification probability maintenance discriminant analysis algorithm, and performing self-adaptive real-time flow data fusion processing on the extracted composite detection signal defect features;
and 4, step 4: the central server submits the fusion result generated in the step 3 for a worker to check;
in the step 2, the specific process of the parallel fusion pretreatment is as follows: defining an initial elastic distributed data set RDD: reading composite detection signal data from an external data space when a cluster computing engine Spark program runs into a system, converting the composite detection signal data into cluster computing engine Spark data blocks, and forming an initial elastic distributed data set RDD;
in the step 3, the specific process of establishing the composite detection signal defect feature database is as follows: the defect modeling and inputting module in the central server software is used for depicting and representing the defect characteristics of the composite detection signal, and a multi-dimensional classification model of the defect characteristics of the composite detection signal is established; constructing a composite detection signal defect feature vector matrix according to the multi-dimensional classification model, and simply classifying defect features; simulating large-scale composite detection signal defect characteristic data by using a simulator, and establishing a typical composite detection signal defect characteristic system and a database;
in the step 3, the specific process of extracting the defect characteristics of the composite detection signal is as follows: taking data in a composite detection signal defect feature database as original samples, calculating the classification probability of each sample to obtain a corresponding class center point and classification information, representing the physical distribution of the corresponding samples through the class center point and the classification information, obtaining more effective composite detection signal defect features representing the original samples, and extracting the composite detection signal defect features;
in the step 3, the specific process of performing the adaptive real-time stream data fusion processing on the extracted defect characteristics of the composite detection signal is as follows:
when the m characteristics are fused, the method can be used,m sets of classifiers are required, the probability output of the classifier is noted as pij(class i | input), wherein i represents the i-th class decision tag, and j represents the j-th class of characteristics generated by the j-th group of classifiers as the decision result;
for the test sample, the probability output corresponding to the decision result of the jth group of classifiers is the maximum value of the probability output corresponding to all decision labels of the group of classifiers, and is recorded as pj：
Then merge weight wjComprises the following steps:
the extracted defect characteristics of the composite detection signal are used as test samples and input into a k-class PSVM classifier, k classes of defect characteristic signals to be trained and classified are set, and m types of defect characteristics of the extracted composite detection signal are set;
respectively inputting m different defect characteristics to be trained into m groups of classifiers, and training according to a tree structure, wherein each group of classifiers comprises 1 PSVM; when the input of the PSVM is more than two types, uniformly dividing the category labels into two categories for judgment until the input of the PSVM at the bottom layer is the two categories, and finishing the tree structure training;
respectively inputting m different defect characteristics to be tested into m groups of classifiers to obtain probability output p of all the classifiersij(class i | input), m groups of classifiers have m probability outputs, and then the fusion weight w corresponding to each group of classifiers is calculatedj；
Weighting the probability output corresponding to each decision label in each group of classifiers, wherein the weighting result is recorded as f (x), f (x) represents the class weight, and the calculation formula is as follows:
and outputting the judgment label corresponding to the maximum weighting result as a fusion result.
2. The data fusion method for blind engineering tracking audit according to claim 1, wherein in step 1, the nondestructive testing sensors comprise infrared thermal imaging equipment, an ultrasonic detector and an eddy current sensor, which are respectively used for detecting the near-surface image, the internal large-depth defect, the defect of a specific steel plate, the diameter and the number of steel bars in the blind engineering.
3. The data fusion method for covert engineering tracking audit according to claim 1, wherein the specific process of the step 4 is as follows: and the central server utilizes a cluster computing engine Spark application program to perform conversion operation on the initial elastic distributed data set RDD to form a new elastic distributed data set RDD, triggers a cluster computing engine Spark driver and submits a fusion result generated by the classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110141287.XA CN112836274B (en) | 2021-02-02 | 2021-02-02 | Data fusion method for tracing and auditing hidden engineering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110141287.XA CN112836274B (en) | 2021-02-02 | 2021-02-02 | Data fusion method for tracing and auditing hidden engineering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112836274A CN112836274A (en) | 2021-05-25 |
| CN112836274B true CN112836274B (en) | 2022-02-15 |
Family
ID=75931529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110141287.XA Active CN112836274B (en) | 2021-02-02 | 2021-02-02 | Data fusion method for tracing and auditing hidden engineering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112836274B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115018456B (en) * | 2022-06-01 | 2023-06-20 | 常州机电职业技术学院 | Data fusion system and method for hidden engineering tracking audit |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830840A (en) * | 2018-05-31 | 2018-11-16 | 江苏华鹰光电科技有限公司 | A kind of active intelligent detecting method of circuit board defect and its application |
| WO2020033169A1 (en) * | 2018-08-07 | 2020-02-13 | Magik Eye Inc. | Baffles for three-dimensional sensors having spherical fields of view |
| CN111985126B (en) * | 2020-07-16 | 2024-04-05 | 河海大学 | Nondestructive testing method for multiple defects in concrete concealed engineering |
-
2021
- 2021-02-02 CN CN202110141287.XA patent/CN112836274B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112836274A (en) | 2021-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106529410B (en) | Monitoring video-based gray haze diffusion path drawing and source determining method | |
| CN106355166B (en) | Dust-haze diffusion path drawing and source determining method based on monitoring video and remote sensing image | |
| CN110717464B (en) | Intelligent railway roadbed disease identification method based on radar data | |
| CN108897975A (en) | Coalbed gas logging air content prediction technique based on deepness belief network | |
| CN102158486A (en) | Method for rapidly detecting network invasion | |
| CN112948932A (en) | Surrounding rock grade prediction method based on TSP forecast data and XGboost algorithm | |
| CN112507479B (en) | Oil drilling machine health state assessment method based on manifold learning and softmax | |
| CN109387565A (en) | A method of brake block internal flaw is detected by analysis voice signal | |
| Xu et al. | Rail defect detection method based on recurrent neural network | |
| CN115774851B (en) | Method and system for detecting internal defects of crankshaft based on hierarchical knowledge distillation | |
| CN110533007A (en) | A kind of vehicle-mounted strain of bridge influences the identification of line feature intelligent and extracting method | |
| CN113009447A (en) | Road underground cavity detection early warning method based on deep learning and ground penetrating radar | |
| CN115654381A (en) | Water supply pipeline leakage detection method based on graph neural network | |
| CN111275744B (en) | Non-contact vibration frequency measurement method based on deep learning and image processing | |
| CN112836274B (en) | Data fusion method for tracing and auditing hidden engineering | |
| Su et al. | New particle formation event detection with Mask R-CNN | |
| CN110889207A (en) | System combination model credibility intelligent evaluation method based on deep learning | |
| Chen et al. | A mixed samples-driven methodology based on denoising diffusion probabilistic model for identifying damage in carbon fiber composite structures | |
| CN116466408B (en) | Artificial neural network superbedrock identification method based on aeromagnetic data | |
| Bi et al. | A fault diagnosis algorithm for wind turbine blades based on bp neural network | |
| CN111651930B (en) | Extreme learning machine-based flow field vortex region detection method | |
| CN111353550A (en) | Method for detecting surface defects of structure based on random structure forest | |
| Yan et al. | YOLOV4-Based Wind Turbine Blade Crack Defect Detection | |
| JP2021108031A (en) | Model generation device, vehicle simulation system, model generation method, vehicle simulation method and computer program | |
| Yan et al. | Water Quality Detection Based on FCN and Embedded System |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |